diff options
| author | Paul Sokolovsky <pfalcon@users.sourceforge.net> | 2014-01-07 20:12:26 +0200 | 
|---|---|---|
| committer | Paul Sokolovsky <pfalcon@users.sourceforge.net> | 2014-01-08 17:38:38 +0200 | 
| commit | 5b15daf6d1f41c166d8c94677663d2d5f91538d1 (patch) | |
| tree | ffeb8739daf9242f6c1f8e25cf1519e79b0a2c58 /py/obj.h | |
| parent | ea9e441a75d7ea6633b0ea95b21667c48a6f7b6a (diff) | |
Add support for stream and buffer protocols.
Stream protocol is abstraction of serial I/O. Buffer protocol is
abstraction of random-access I/O. These protocols are defined down
to C level, to allow generic, while still efficient algorithms
to be coded in C (like, buffered transfer between 2 stream objects,
saving/loading of buffer object to/from stream, etc). (Note that CPython
define buffer protocol on C level, but apparently not stream protocol).
Diffstat (limited to 'py/obj.h')
| -rw-r--r-- | py/obj.h | 41 | 
1 files changed, 41 insertions, 0 deletions
@@ -76,6 +76,41 @@ typedef struct _mp_method_t {      mp_const_obj_t fun;  } mp_method_t; +// Buffer protocol +typedef struct _buffer_info_t { +    // if we'd bother to support various versions of structure +    // (with different number of fields), we can distinguish +    // them with ver = sizeof(struct). Cons: overkill for *micro*? +    //int ver; // ? + +    void *buf; +    machine_int_t len; + +    // Rationale: have array.array and have SIMD operations on them +    // Cons: users can pass item size to processing functions themselves, +    // though that's not "plug&play" +    // int itemsize; + +    // Rationale: to load arbitrary-sized sprites directly to LCD +    // Cons: a bit adhoc usecase +    // int stride; +} buffer_info_t; +#define BUFFER_READ  (1) +#define BUFFER_WRITE (2) +#define BUFFER_RW (BUFFER_READ | BUFFER_WRITE) +typedef struct _mp_buffer_p_t { +    machine_int_t (*get_buffer)(mp_obj_t obj, buffer_info_t *bufinfo, int flags); +} mp_buffer_p_t; + +// Stream protocol +typedef struct _mp_stream_p_t { +    // On error, functions should return -1 and fill in *errcode (values are +    // implementation-dependent, but will be exposed to user, e.g. via exception). +    machine_int_t (*read)(mp_obj_t obj, void *buf, machine_uint_t size, int *errcode); +    machine_int_t (*write)(mp_obj_t obj, const void *buf, machine_uint_t size, int *errcode); +    // add seek() ? +} mp_stream_p_t; +  struct _mp_obj_type_t {      mp_obj_base_t base;      const char *name; @@ -90,6 +125,12 @@ struct _mp_obj_type_t {      mp_fun_1_t getiter;      mp_fun_1_t iternext; +    // Alternatively, pointer(s) to interfaces to save space +    // in mp_obj_type_t at the expense of extra pointer and extra dereference +    // when actually used. +    mp_buffer_p_t buffer_p; +    mp_stream_p_t stream_p; +      const mp_method_t *methods;      /*  | 
